Re: [CAD_CAM_EDM_DRO] Re: [Lost Steps

beer@... wrote:

> On 23 Aug, CAD_CAM_EDM_DRO@egroups.com wrote:
> >
> > If you don't want to have a system plagued with noise problems, do
> > the ground wiring right from the beginning. Right means use the
> > concept of a "holy point ground". This means ALL ground wires are
> > returned to a single point; power supply ground, motor drive ground,
>
> > parallel port ground, in short if it has a ground wire, take it to
> > that point! If shields are used, take them there too.
>
> Agree wholeheartedly with this - actually, there is really nothing to
> agree upon, this is fact, plain and simple.
>
> But to add a couple of refresher points to clarify this ...
>
> 1. "Wire" is a theoretical thing, that does not exist in nature. The
> stuff we call wire, the stuff we use to "wire things up", is always a
> complex thing, exhibiting resistance, inductance, and frequently
> capacitance.
>
> 2. There is only ONE ground point in a circuit. All other parts of
> the
> circuit connected to this point are not quite at ground, because of #1
>
> above.
>
> 3. The engineer who designs any particular circuit hopefully takes
> this
> real world into account. Engineers who design circuits with high gain
>
> and/or high currents MUST take this into account, again, because of
> #1.
> This board has both high gain and high currents.
>
> Which leads us back to the single point ground. All of the various
> ground points in the system MUST go back to a single point, and that
> single point should be the negative terminal of the main filter cap.

Up to this point, you are definitely right. But, here, you may be
making a mistake.
Long wires running from the common point to the driver grounds will
develop
a voltage. These drivers need to be at as close to the same voltage as
the common
point, because they are also used as a reference for the step and
direction
pulses coming in from the computer. So, in this case, the common point
should be something like a big bus bar running as close as possible to
the
ground terminals on the drivers. This would be my solution in this
particular
case. I have used similar techniques in the past with good results. I
have
the same thing in my PWM servo amp setup, and it works fine. The whole
idea
is to minimize the loop area of the connections, as that reduces
inductance.

>
> There is an argument that can be made that the negative terminal of
> a
> bridge rectifier is more negative than the negative terminal of the
>
> filter cap, but a close examination would show that under high
> instantaneous current conditions, that point might actually be
> positive with respect to the cap. At any rate, bringing ALL
> ground connections back to the cap's negative terminal will solve
> grounding problems.
>
> That means that bringing a "wire" from the power supply end of the
> board as well as a "wire" from the logic end of the board back to the
> filter cap. Also, the "ground" wire(s) from the input connection (
> from
> the parallel port ) must also go back to this single point.
>
>
> I don't think that noise is the problem. With all due respect to
> those
> who have suggested cures for noise problems ( and most of those cures
> can't hurt ), I think this is the wrong track. The inputs to this
> board are active low, and are pulled high by a resistor ( 1k? ). It
> would take a HUGE amount of noise to force that pullup to ground ( OK,
>
> <1/2 Vcc ) and thereby create a step pulse, and none of that noise
> would exist at any level with the unit in steady-state, i.e., not
> stepping.

Well, it is quite easy to prove you are wrong, here. If I was on site
with an
oscilloscope, I could easily show you several volts of noise in this
system,
I'm sure. I have seen it plenty of times. MOSFETs switch from fully on
to
off (or vice versa) in about 40 nS if driven hard, and in 100 nS or so
if
driven more softly. Either way, it only takes micro Henries of
inductance
to generate volts when the di/dT is in the area of 50 x 10^6
Volts/Second!
In fact, if not for parasitic capacitances that slow things down, a
di/dT of
50 x 10^6 V/Sec, across one uH inductance, will generate 50 Volts!
And, a foot of wire (no matter what the gauge) will exhibit several
micro Henries
of inductance, easily. That is why the common point MUST be close to
the
source of the current pulses, and the wires must be short.

Jon